CWL workflow complied to include a subworkflow with an `outputs` applet fails to run successfully

[ThomasHickman]

If you have a workflow as follows:

cwlVersion: v1.2
class: Workflow
inputs: []
requirements:
  SubworkflowFeatureRequirement: {}
steps:
    - id: one
      in: []
      out: [output]
      run: one.cwl
    - id: two
      in: []
      out: [output]
      run: two_wrapper.cwl
outputs:
    output1:
      type: File
      outputSource: one/output
    output2:
      type: File
      outputSource: two/output

one.cwl:

cwlVersion: v1.2
class: CommandLineTool
inputs: []
baseCommand:
    - bash
    - -c
    - "echo something > output"
outputs:
    - id: output
      type: File
      outputBinding:
        glob: output

two_wrapper.cwl:

cwlVersion: v1.2
class: Workflow
inputs: []
steps:
    - id: two
      in: []
      out: [output]
      run: two.cwl
outputs:
    output:
      type: File
      outputSource: two/output

two.cwl:

cwlVersion: v1.2
class: CommandLineTool
inputs: []
baseCommand:
    - bash
    - -c
    - "echo something > output"
outputs:
    - id: output
      type: File?
      outputBinding:
        glob: output

`cwlpack` workflow

    "class": "Workflow",
    "cwlVersion": "v1.2",
    "id": "workflow.cwl",
    "inputs": [],
    "outputs": [
        {
            "id": "output1",
            "outputSource": "one/output",
            "type": "File"
        },
        {
            "id": "output2",
            "outputSource": "two/output",
            "type": "File"
        }
    ],
    "requirements": [
        {
            "class": "SubworkflowFeatureRequirement"
        }
    ],
    "steps": [
        {
            "id": "one",
            "in": [],
            "out": [
                "output"
            ],
            "run": {
                "baseCommand": [
                    "bash",
                    "-c",
                    "echo something > output"
                ],
                "class": "CommandLineTool",
                "cwlVersion": "v1.2",
                "id": "workflow.cwl@step_one@one.cwl",
                "inputs": [],
                "outputs": [
                    {
                        "id": "output",
                        "outputBinding": {
                            "glob": "output"
                        },
                        "type": "File"
                    }
                ],
                "requirements": []
            }
        },
        {
            "id": "two",
            "in": [],
            "out": [
                "output"
            ],
            "run": {
                "class": "Workflow",
                "cwlVersion": "v1.2",
                "id": "workflow.cwl@step_two@two_wrapper.cwl",
                "inputs": [],
                "outputs": [
                    {
                        "id": "output",
                        "outputSource": "two/output",
                        "type": "File"
                    }
                ],
                "requirements": [],
                "steps": [
                    {
                        "id": "two",
                        "in": [],
                        "out": [
                            "output"
                        ],
                        "run": {
                            "baseCommand": [
                                "bash",
                                "-c",
                                "echo something > output"
                            ],
                            "class": "CommandLineTool",
                            "cwlVersion": "v1.2",
                            "id": "two_wrapper.cwl@step_two@two.cwl",
                            "inputs": [],
                            "outputs": [
                                {
                                    "id": "output",
                                    "outputBinding": {
                                        "glob": "output"
                                    },
                                    "type": [
                                        "null",
                                        "File"
                                    ]
                                }
                            ],
                            "requirements": []
                        }
                    }
                ]
            }
        }
    ]
}

This generates the following applets:

Note: there are two outputs applets, the inner workflow_cwl_step_two_two_wrapper_cwl_outputs applet created by two_wrapper.cwl narrowing the output type from File? in two.cwl to File in two_wrapper.cwl.

When running this, two_wrapper's outputs step fails as seen below:

with the error message:

Environment: Map(two/output -> (TOptional(TFile),VFile(dx://file-GYqV19QJKB2b8pXF24xJ1xBb::/output,Some(output),None,Some(sha1$50a4e988380c09d290acdab4bd53d24ee7b497df),Some(10),Vector(),None,None)))
Evaluating workflow outputs
Evaluating output parameters:
  (output1,WorkflowOutputParameter(Some(Identifier(Some(file:/null),workflow.cwl/output1)),None,None,CwlFile,Vector(),None,false,Vector(Identifier(Some(file:/null),one/output)),None,None),CwlFile)
  (output2,WorkflowOutputParameter(Some(Identifier(Some(file:/null),workflow.cwl/output2)),None,None,CwlFile,Vector(),None,false,Vector(Identifier(Some(file:/null),two/output)),None,None),CwlFile)
[error] failure executing Workflow action 'Outputs'
java.lang.Exception: cannot coerce VNull to TFile
	at dx.core.ir.Value$.coerceTo(Value.scala:311)
	at dx.executor.cwl.CwlWorkflowExecutor.$anonfun$evaluateOutputs$3(CwlWorkflowExecutor.scala:329)
	at scala.collection.immutable.Vector1.map(Vector.scala:1872)
	at scala.collection.immutable.Vector1.map(Vector.scala:375)
	at dx.executor.cwl.CwlWorkflowExecutor.evaluateOutputs(CwlWorkflowExecutor.scala:283)
	at dx.executor.WorkflowExecutor.evaluateOutputs(WorkflowExecutor.scala:156)
	at dx.executor.WorkflowExecutor.apply(WorkflowExecutor.scala:897)
	at dx.executor.BaseCli.dispatchCommand(BaseCli.scala:103)
	at dx.executor.BaseCli.main(BaseCli.scala:137)
	at dxExecutorCwl.MainApp$.delayedEndpoint$dxExecutorCwl$MainApp$1(Main.scala:27)
	at dxExecutorCwl.MainApp$delayedInit$body.apply(Main.scala:26)
	at scala.Function0.apply$mcV$sp(Function0.scala:39)
	at scala.Function0.apply$mcV$sp$(Function0.scala:39)
	at scala.runtime.AbstractFunction0.apply$mcV$sp(AbstractFunction0.scala:17)
	at scala.App.$anonfun$main$1(App.scala:76)
	at scala.App.$anonfun$main$1$adapted(App.scala:76)
	at scala.collection.IterableOnceOps.foreach(IterableOnce.scala:563)
	at scala.collection.IterableOnceOps.foreach$(IterableOnce.scala:561)
	at scala.collection.AbstractIterable.foreach(Iterable.scala:926)
	at scala.App.main(App.scala:76)
	at scala.App.main$(App.scala:74)
	at dxExecutorCwl.MainApp$.main(Main.scala:26)
	at dxExecutorCwl.MainApp.main(Main.scala)

implying that it's trying to collect outputs for the entire workflow, not the two workflow.

I've had a bit of a delve into the dxCompiler source code to see what might be going on. My theory is that:

1. Prior to CwlWorkflowExecutor.scala:329 (mentioned in the exception), we find that the output parameters are determined from the workflow property in the CwlWorkflowExecutor class here:
   https://github.com/dnanexus/dxCompiler/blob/develop/executorCwl/src/main/scala/dx/executor/cwl/CwlWorkflowExecutor.scala#L271
2. The workflow that it fetches here seems to be from the CwlWorkflowExecutor.create function. In this function, the following bit of code executes this to find the appropriate workflow:
   https://github.com/dnanexus/dxCompiler/blob/develop/executorCwl/src/main/scala/dx/executor/cwl/CwlWorkflowExecutor.scala#L78-L82
3. In this bit of ciode OriginalName of the outputs applet is "${wfName}_outputs", as can be seen here:
   https://github.com/dnanexus/dxCompiler/blob/develop/compiler/src/main/scala/dx/translator/cwl/ProcessTranslator.scala#L748
4. CwlWorkflowExecutor.create therefore follows:
   https://github.com/dnanexus/dxCompiler/blob/develop/executorCwl/src/main/scala/dx/executor/cwl/CwlWorkflowExecutor.scala#L98-L103
   and considers the wrong workflow

[Gvaihir]

Thanks for the report! At the moment we have a very limited support for CWL, so it may take significant time before our team is able to prioritize it.